Microsoft claims it has built a multi-agent AI system capable of diagnosing medical conditions up to four times more accurately than human doctors, according to The Verge. Published Thursday in the journal Science, tests showed the technology handled complex patient scenarios from emergency departments and case reports with striking precision.
Key takeaways
- Microsoft created an AI tool named the Diagnostic Orchestrator that reasons through medical histories.
- The system achieved around 83 percent accuracy on cases from The New England Journal of Medicine.
- Researchers tested the technology against generalist physicians with and without reference materials.
- Microsoft plans to begin hospital testing with the Beth Israel Deaconess Medical Center in Boston.
How does the microsoft diagnostic AI work?
Designed to function like a panel of medical experts, the Diagnostic Orchestrator uses multiple AI agents to reason through symptoms, order tests, and arrive at a diagnosis. The Verge reports that the system is powered by OpenAI's GPT-5 and mimics the exact process human doctors use to investigate medical cases.
During trials on emergency department cases from the Beth Israel Deaconess Medical Center in Boston, the system scored 80 percent accuracy. Human physicians averaged 47 percent on the same patient admissions data. Also, the AI reached conclusions at a lower estimated cost than its human counterparts.
What is the diagnostic Orchestrator?
Microsoft compiled its test cases into a benchmark called SDBench, which the company intends to open-source. Mustafa Suleyman, Microsoft AI CEO, told The Verge that the technology represents a major milestone. Even so, the system has only been evaluated on past case reports rather than live patients.
Dr. Dominic King, the company's ai health lead and coauthor of the study, noted that clinical validation remains necessary before a wider rollout. Real-world hospital trials with the Boston medical center are slated to start in the next couple of months.
What happens next for medical superintelligence?
The research forms part of a broader corporate push toward medical superintelligence. Suleyman predicts that systems matching human performance in this field will arrive within the next couple of years. Commercialization plans remain unannounced as executives weigh high regulatory and compliance requirements.
While healthcare deployments will accelerate next year, human clinicians will retain a central role in patient care. Developers hope the technology will remove administrative burdens and give doctors more time to spend directly with patients.
Frequently asked questions
It is an AI-powered tool that uses multiple agents to reason through symptoms, decide what tests to order, and arrive at a diagnosis, mimicking human doctors.
The system achieved around 83 percent diagnostic accuracy on cases from The New England Journal of Medicine and 80 percent on emergency department cases.
The findings were published Thursday in the academic journal Science, alongside additional articles in BMJ Digital Health & AI.
The Diagnostic Orchestrator is powered by OpenAI's GPT-5, though it was initially developed when GPT-4.1 was the most advanced available model.
Testing in real hospital settings with the Beth Israel Deaconess Medical Center in Boston will likely get underway in the next couple of months.
Based on reporting by The Verge. Read the original article
